AI for Science

We investigate the magnetic ground states and transverse spin excitations of bilayer and trilayer nickelates within a multi-orbital itinerant framework. For the bilayer system, although Hartree-Fock calculations slightly favor a double-stripe order, the calculated excitation spectrum of the single-stripe state, characterized by an anisotropic low-energy cone at $Q_{\text{BL}}$ and isotropic high-energy excitations near $Γ$, exhibits good qualitative agreement with recent RIXS and neutron scattering experiments. We further identify mirror-even optical interlayer modes at $Q_{\text{BL}}$ whose energies match the mirror-odd modes at $Γ$. For the trilayer system, both mirror-odd and mirror-even spin-density-wave states can be stabilized near $Q_{\text{TL}}$, with the mirror-odd state lower in energy in the parameter regime studied. The mirror-odd state hosts an additional nearly gapless mode dominated by the middle layer, while the mirror-even state contains only one acoustic branch together with two gapped optical modes. Comparison with available RIXS data favors the mirror-odd spin-density-wave scenario. Our results show that magnetic excitations provide a sensitive probe of the magnetic order and support a common itinerant origin of magnetism in multilayer nickelates.

Amorphous silicon (a-Si) is understood to be the canonical continuous random network material, ideally defined by fully fourfold coordination. Here, we show that a defect-free ('ideal') model of a-Si from machine-learning-driven molecular-dynamics simulations [L. A. M. Rosset et al., Nat. Commun. 16, 2360 (2025)], subsequently evaluated with hybrid-level density-functional theory computations, can accurately reproduce the experimentally observed electronic bandgap. We compare this model with one resulting from the Wooten-Winer-Weaire (WWW) bond-switching approach and with other recent approximants to ideal a-Si. More broadly, our work provides a platform for studies of band tails, optical properties, and transport in a-Si.

The topological heavy-fermion model provides a minimal framework for describing the coexistence of localized moments and itinerant Dirac electrons in magic-angle twisted bilayer graphene (MATBG). Several analytical and numerical methods have been applied to this model; however, whether they provide a realistic description of MATBG remains incompletely understood. In this work, we develop an Hubbard operator approach that incorporates non-local correlations beyond the single-site limit. We benchmark the approximate calculations against numerically exact determinant quantum Monte Carlo simulations of a lattice-regularized model. We show that commonly used local approximations, such as Hubbard-I, fail to capture the coupling between localized and itinerant degrees of freedom, leading to incorrect spectral properties in the local-moment regime. In contrast, the Hubbard operator method provides a controlled description of both correlation functions and spectral features over a regime of parameters, in good agreement with exact numerical methods.

We use density functional theory and model Hamiltonians to reveal large spin splitting of bands localized at low-symmetry, ferromagnetic surfaces of bulk antiferromagnets (AFMs). There is great interest in finding new material platforms combining the robustness and ultrafast dynamics of AFMs with large, functional spin splitting which is often restricted to ferromagnets. Here, we show that a subset of AFM surfaces which have symmetry-allowed magnetization can host large spin splitting via bulk degeneracy lifting of sublattice-resolved exchange splittings. Using model Hamiltonians, we show that the spin splitting is maximized for two ferromagnetic surface motifs: terminations with single uncompensated magnetic sublattices, and two-sublattice surfaces whose sublattices are magnetically and electronically compensated in the bulk, but acquire distinct crystal field environments via surface truncation. The latter case can yield FM-like spin splitting magnitudes while also having vanishingly small uncompensated magnetization. In contrast, when surface magnetization arises from relativistic canting on symmetry-connected sublattices, the spin splitting is expected to be small. We confirm these predictions with first-principles calculations of $\mathrm{Cr_2O_3}$ and $\mathrm{FeF_2}$, finding splittings from $\sim10\mathrm{meV}$-$\sim1\mathrm{eV}$ depending on the surface in question. Our findings point to intrinsic surface symmetry breaking as a route to large, functional spin splitting in an expanded range of AFM materials.

Accurate modeling of dielectric screening in van der Waals (vdW) heterostructures is essential for predicting photonic and optoelectronic properties - yet conventional first-principles methods are often hindered by incommensurate lattices and prohibitive computational costs. In this work, we introduce the microscopic Quantum Electrostatic Heterostructure (mQEH) method. mQEH employs a hierarchical and systematically improvable basis set to describe potentials and induced densities, eliminating the need for arbitrary geometric cutoffs and ensuring accurate screening descriptions at all length scales. The mQEH method is combined with a layer projected Bethe-Salpeter Equation (BSE) to enable calculations of optical spectra of experimentally relevant lattice-mismatched vdW heterostructures. Applying the mQEH-BSE framework to a series of transition-metal dichalcogenide (TMD) heterobilayers, we obtain absorption spectra and momentum-indirect exciton energies in excellent agreement with experiment. The framework provides a computationally efficient route to predictive modeling and design of vdW heterostructures with tailored optical properties.

Sub-bandgap optical absorption in AlN between 2 eV and 4 eV is widely observed, but its microscopic origin remains contested. Using photo-induced electron paramagnetic resonance (photo-EPR) and optical absorption spectroscopy on the same samples, we demonstrate a correlation between this absorption band and the neutral charge state of substitutional carbon on the nitrogen site (C$_N$). Hybrid functional calculations of the optical absorption spectra show that a transition involving C$_N$ and the valence band occurs near 3.3 eV, which agrees well with a peak identified within the measured optical absorption between 2 eV and 4 eV. This conclusion requires the combined ability to manipulate the charge state of carbon using photo-EPR and to use first-principles calculations of the absorption line shape that account for the dispersion of the valence band and the energy dependence of the optical matrix elements.

The growth of a thin film of a high-entropy AlCoCuFeNi alloy on a silicon (100) substrate was studied using molecular dynamics modeling. The simulation was carried out using the embedded atom model to describe the interactions among Al, Co, Cu, Ni, and Fe atoms. The interaction between Al, Co, Cu, Fe, Ni atoms and the Si substrate was modeled using the Lennard-Jones potential, while the interaction between silicon atoms was described using the Stillinger-Weber potential. The total simulation time was 50 ns. It was found that small clusters were formed at the first stage of deposition and that crystallization started after approximately 5 ns of simulation, when the characteristic cluster size was about 2 nm. At the end of the simulation, after 50 ns of modeling, the simulated film contained face-centered cubic, body-centered cubic, hexagonal close-packed, and amorphous phases. Analysis of the radial distribution function made it possible to determine nearest-neighbor distances and estimate the lattice parameters of these phases.

Identification of the particle interaction potential is a challenging and important task in dusty plasma, colloids, and smart materials as it allows the characterization of structure formation and helps predict phase transitions. With the advent of machine learning methods, this interaction can be extracted from particle position data, leading to a generalizable expression which is applicable in different systems. Methods such as sparse regression aim to provide a physically interpretable model that can generalize well, while avoiding unnecessary complexity due to overfitting. In this work, we present the use of the Sparse Identification of Nonlinear Dynamics (SINDy) with the weak formulation to learn equations of motion for noisy data from simple simulations of two dust particles interacting with a Yukawa (shielded Coulomb) potential. The application of these methods to experimental dusty plasma data is discussed, particularly in the case of simulation data and glass box experiments in RF discharge gravity environments and DC discharge microgravity environments, such as the Plasmakristall-4 (PK-4) experiment.

While correlated phenomena of flat bands have been extensively studied in twisted systems, the ordered states that emerge from interactions in the intrinsic flat bands of kagome lattice materials remain largely unexplored. The newly discovered kagome metal CsCr3Sb5 offers a unique and rich platform for this research, as its multi-orbital flat bands at the Fermi surface result in a complex interplay of pressurized superconductivity, antiferromagnetism, a structural phase transition, and density wave orders. Here, using ultrafast optical techniques, we provide strong spectroscopic evidence for a charge density wave transition in CsCr3Sb5, resolving previous ambiguities. Crucially, we identify rotational symmetry breaking that manifests as a three-state Potts-type nematicity. Our elastoresistance measurements directly demonstrate the electronic origin of this order, as the rotational-symmetry-breaking E2g component of the elastoresistance shows a divergent behaviour around the transition temperature. This exotic nematicity results from the lifting of degeneracy of the multi-orbital flat bands, akin to phenomena seen in certain iron-based superconductors. Our study pioneers the investigation of ultrafast dynamics in flat-band systems at the Fermi surface, offering new insights into the interactions between multiple elementary excitations in strongly correlated systems.

Variational infinite matrix product state (iMPS) calculations usually make Hamiltonians with algebraically decaying interactions compatible with standard MPO algorithms by first replacing the target Hamiltonian with a finite-pole sum-of-exponentials surrogate, thereby introducing a Hamiltonian-representation residual. We formulate the fixed-$D$ variational energy without introducing such a surrogate. For a fixed finite-$D$ MPS, the algebraic tail can be summed directly through the connected transfer matrix: the tail $e^{\mathrm{i} Qr}/r^α$ is represented by the matrix function $F_{α,Q}(\widetilde{T}_A)$, with $F_{α,Q}(z)=\operatorname{Li}_α(e^{\mathrm{i} Q}\,z)/z$. We evaluate the resulting matrix-function action using a Krylov method and obtain stable gradients by combining a Fréchet adjoint with implicit fixed-point differentiation. Benchmarks on long-range free fermions and the inverse-square Heisenberg family, including the Haldane--Shastry point, validate the transfer-matrix-function formulation. A long-range Ising-chain calculation illustrates a practical consequence of avoiding a finite-pole Hamiltonian representation. At a fixed, independently known critical field, finite-pole surrogate Hamiltonians can bias a critical diagnostic away from criticality, whereas the matrix-function calculation retains the expected critical signatures of the target algebraic Hamiltonian.

We consider protocols for control of dipolar Bose-Einstein condensates where the critical role is played by the long-range anisotropic interatomic magnetic dipole-dipole interaction. The phase diagram of such a condensate has been explored theoretically and experimentally with certain values of the interatomic scattering length corresponding to superfluid and supersolid phases, where supersolidity appears as a modulation in the ground state density. Preparation of this modulated ground state is challenging, since excitations appear as a result of a finite-time evolution required to produce qualitative changes in the wavefunction density. To solve this problem we consider the time-dependent control of a dipolar Bose-Einstein condensate using shortcuts to adiabaticity techniques, concentrating on design of the time-dependent scattering length, a parameter of the system easily tunable by contemporary experiments. The first technique is the variational approach based on the Euler-Lagrange equations for a separable ansatz describing the evolution of the superfluid state. Secondly, we study the transition from superfluid to supersolid using a direct optimization protocol. We discuss the fidelity of the developed protocols in terms of the evolution time.

Statistical methods are essential for understanding thermodynamic systems with many degrees of freedom. For systems in equilibrium, a very useful method is that of correlation functions, which establish a correlation between a field phi(x), which depends on the spatial position x, and the same field evaluated at another position, phi(x0). Fisher [Journal of Mathematical Physics 5, 944322 (1964)] introduced the autocorrelation function for fluctuations of the order parameter, which has been an important mathematical tool for understanding second-order phase transitions in equilibrium. However, his analysis is restricted to a Euclidean space of dimension d, and an exponent eta is introduced to correct the spatial behavior of the correlation function at T = Tc. In a recent work, Lima et al. [Phys. Rev. E 110, L062107 (2024)] demonstrated that a modern fractional differential analysis is necessary for a complete description of the correlation function at Tc. In this study, we highlight the deep connection among scaling behavior, critical exponents, and fractal geometry. Our results provide a unified geometric interpretation of critical exponents and fractal dimensions, broadly applicable to thermodynamic phase transitions. However, the approach does not apply to topological phase transitions, which lack local order parameters and the associated scale-invariant fractal geometry. We verify its predictions for several cornerstone thermodynamic models: the Ising, Potts, XY, and Heisenberg systems.

Characterising the universal aspects of non-equilibrium quantum many-body dynamics is one of the key goals of this century's physics research. Progress, however, is hindered by the lack of general theoretical frameworks for studying interacting quantum matter far from equilibrium. A recent breakthrough has been the realization that several key non-equilibrium quantities, such as the rate of growth of entanglement or the fluctuations of conserved charges within finite subsystems, can be related to equilibrium properties through a space-time duality that effectively exchanges the roles of space and time. This observation effectively enables the study of non-equilibrium phenomena using tools and concepts borrowed from equilibrium statistical mechanics and thermodynamics. A first proof of principle of this framework, dubbed space-time duality approach (SDA), was provided by interacting integrable systems, where thermodynamic properties can often be characterized exactly, while dynamical quantities typically remain beyond analytical reach. Subsequent developments, however, revealed that the SDA suffered from an intrinsic ambiguity, restricting its applicability to homogeneous quenches and to charge fluctuations arising from symmetric initial states. Here we resolve this ambiguity from first principles and derive closed-form predictions for entanglement growth and charge fluctuations after general quantum quenches. We benchmark our results against the exact analytical solution of the Rule 54 quantum cellular automaton and extensive TEBD simulations of the XXZ chain. Moreover we show that, when specialised to the entanglement entropy, our framework naturally reproduces the predictions of the quasiparticle picture.

When stochastic field dynamics are cast into a path-integral formulation, perturbation theory becomes systematic but the resulting expansion quickly grows combinatorially large. The setting targeted here includes multi-component, multi-dimensional fields with matrix propagators, tensor-valued couplings, and non-Gaussian driving noise specified by arbitrary $n$-point cumulants. Wick pairings grow factorially, and component indices must be routed through the tensor-valued vertices. The useful output is not a raw contraction list, but a diagram table: one entry per topology, with multiplicities, coupling sums, signs, and causal constraints resolved. We present sft-wick, an open-source Python package that constructs these diagram tables and computes their integrals numerically. Given an action and an observable, it enumerates topologically distinct Feynman diagrams, derives their algebraic coefficients, and evaluates the resulting diagram integrals from user-supplied response and cumulant functions. The core algorithm enumerates spatial topologies before routing component indices, avoiding contraction-by-contraction Wick expansion. Response-field constraints, including vanishing response-response contractions, the ito prescription, and the absence of causal response loops, are enforced during enumeration. Predictions are validated against direct Langevin simulation, agreeing to within the simulation's statistical noise.

We formulate reheating as a constrained variational problem in the space of equation-of-state histories, rather than attempting to describe it through microscopic models. We introduce a regularized functional framework that identifies reheating histories which extremize a given cosmological observable under minimal physical assumptions. As illustrative applications, we consider prompt gravitational waves, induced gravitational waves, and primordial black holes. We find that different observables select qualitatively different regions of reheating-history space. These examples demonstrate that cosmological observables define distinct extremal directions in reheating-history space and can therefore be used to systematically explore the space of post-inflationary expansion histories.

The rotational dynamics induced by the non-frontal binary collisions of quantum droplets composed of ultracold alkali atoms are analyzed. A theoretical study is presented within the extended Gross-Pitaevskii equation framework, using experimentally feasible conditions. Numerical experiments elucidate a rich landscape of possible topological excitations in the system that are robust towards measurements. The collision of heteronuclear quantum droplets composed of $^{41}$K and $^{87}$Rb atoms in the incompressible regime, gives rise to dynamical instabilities that spontaneously generate topological defects: vortex rings, dislocation lines, and vortices in one species. Their presence depends on the Weber number and the impact parameter. An experimental proposal for vortex detection in both real and Fourier space using interaction ramps is described.

Deep neural networks (DNNs) have achieved remarkable success in scientific computing, yet they often suffer from spectral bias in capturing oscillatory and multiscale behaviors. In this study, we investigate this limitation by examining the failure of shallow ReLU neural networks in fitting high-frequency functions. This observation identifies two important factors in resolving rapid oscillations: the initial slope scale and the distribution of partition points induced by the networks. Motivated by this analysis, we propose RepNN, a reparameterized neural network model with activation ReLU or tanh designed for high-frequency and multiscale problems. The key idea is to reparameterize the weights and biases in the first hidden layer, which enables effective control of the initial slope scale and provides an appropriate distribution of the initial partition points. Furthermore, treating the reparameterized weights and biases as trainable parameters allows the DNN to achieve adaptive frequency scaling during training. In addition, we derive quantitative estimates for the output and slope magnitudes of the reparameterized DNN to guide the initialization of the proposed method. Numerical experiments, including multiscale one- and four-dimensional function approximations, forward and inverse PDE problems in combination with physics-informed neural networks (PINNs), and operator learning for an earthquake problem using real data, demonstrate that RepNN improves the predicted accuracy of vanilla DNNs in capturing highly oscillatory features with slightly additional computational cost. These results indicate that RepNN provides an effective and flexible approach for overcoming spectral bias and applying DNNs to multiscale problems.

In this work, we investigate the impact of fusion-born alpha particles on core turbulence and transport in the ARC tokamak fusion power plant using linear and nonlinear gyrokinetic CGYRO simulations. A significant reduction in ion-scale turbulent heat and particle fluxes is observed in the inner core (r/a $\leq$ 0.5), which is associated with multiscale interactions between fast ion-destabilized modes, zonal flows, and the background turbulence. A nonlinear upshift in the ITG critical gradient is observed in the simulations with fast alphas compared to those with artificially thermalized alphas. The turbulence reduction is found to scale beneficially with alpha particle density and plasma $β_e$, and the radial extent of the turbulence suppression is limited to the volume containing a significant density of fast particles. The suitability of local gyrokinetics and potential impacts of fast ion effects on fusion performance are discussed.

We study dynamical orthogonal (DO) approximations of stochastic differential equations and investigate their long-time behaviour. The DO formulation represents the solution by a low-rank decomposition and leads to a coupled system consisting of an evolution equation on the Stiefel manifold and a reduced stochastic process. We establish the well-posedness of the strong DO system and derive quantitative error estimates between the original stochastic differential equation and its low-rank approximation in the Wasserstein distance. Our main contribution is the analysis of invariant probability measures for the DO dynamics. Under suitable dissipativity, Lipschitz continuity, and non-degeneracy assumptions on the coefficients, we prove the existence of an invariant probability measure for the strong DO system. The proof combines uniform moment estimates, a Krylov--Bogoliubov argument for an associated frozen system, and a Kakutani-Fan-Glicksberg fixed-point theorem to recover the self-consistent dynamics. We further show that the induced low-rank process admits an invariant probability measure and discuss the structure of invariant measures through several illustrative examples. These results provide a rigorous foundation for the use of dynamical low-rank approximations in the approximation of long-time statistical properties of stochastic dynamical systems.

The variational quantum eigen solver (VQE), has been widely used to find the ground state energy of different Hamiltonians with no analytical solutions and are classically difficult to compute. In our work, we have used VQE to identify the phase transition boundary for an infinite order phase transition. We use long-range XXZ (LRXXZ) chain for our study. In order to probe infinite order phase transition, we propose to utilise the ground state energy obtained from VQE. The idea rests on the argument that VQE requires an ansatz circuit; therefore, the accuracy of the VQE will rely on this ansatz circuit. We have designed this circuit such that the estimated ground state energy is sensitive to the phase it is evaluated in. It is achieved by applying the constraint that the net spin remains constant throughout the optimisation process. Consequently, the ansatz works in a certain phase where it gives relatively small random error, as it should, when compared to the error in ground state energy calculations of the other phases, where the ansatz fails. By identifying these changes in the behaviour of the error in ground state energy using VQE, we were able to determine the phase boundaries. Using exact diagonalisation, we also compare the behaviour of the energy gradient and energy gap across both the phase transition boundaries for this model. Further, by increasing the depth of the optimisation circuit, we also accurately evaluate the ground energy of the LRXXZ chain for the value of coupling constant, J equal to -1

Estimating molecular ground-state energies is a central application of quantum computing, requiring both the preparation of accurate quantum states and efficient energy readout. Understanding the effect of hardware noise on these experiments is crucial to distinguish errors that have low impact, errors that can be mitigated, and errors that must be reduced at the hardware level. We ran a state preparation and energy measurement protocol on an ion-trap quantum computer, without any non-scalable off-loading of computational tasks to classical computers, and show that leakage errors are the main obstacle to chemical accuracy. More specifically, we apply adiabatic state preparation to prepare the ground state of a six-qubit encoding of the H3+ molecule and extract its energy using a noise-resilient variant of iterative quantum phase estimation. Our results improve upon the classical Hartree-Fock energy. Analyzing the effect of hardware noise on the result, we find that while coherent and incoherent noise have little influence, the hardware results are mainly impacted by leakage errors. Absent leakage errors, noisy numerical simulations show that with our experimental settings we would have achieved close to chemical accuracy, even shot noise included. These insights highlight the importance of targeting leakage suppression in future algorithm and hardware development.

Granular creep is the slow, sub-yield movement of constituents in a granular packing due to the disordered nature of its grain-scale interactions. Despite the ubiquity of creep in disordered materials, it is still not understood how to best predict the creep-to-failure regime based on the forces and interactions among constituents. To address this gap, we perform experiments to explore creep and failure in quasi two-dimensional piles of photoelastic disks, allowing the quantification of both grain movements and grain-scale contact force networks. Through controlled external disturbances, we investigate the emergence and evolution of grain rearrangements, force networks, and voids to illuminate signatures of creep and failure. Surprisingly, the force chain structure remains dynamic even in the absence of observable particle motion. We find that shifts in force chains provide an indication to larger, avalanche-scale disruptions. We connect these force signatures with the geometry of the voids in the pile. Overall, our novel experiments and analyses deepen our mechanical and geometric understanding of the creep-to-failure transition in granular systems.

The Doyle-Fuller-Newman (DFN) model resolves internal electrochemical states in lithium-ion batteries with high fidelity. However, the numerical solution of its governing equations is computationally prohibitive for real-time deployment, limiting scalability from individual cells to pack and fleet-scale applications. While machine learning surrogates can substantially reduce inference latency through GPU acceleration, most existing approaches learn solution approximations tied to specific operating conditions rather than learning generalizable state-evolution dynamics. This work presents a systematic comparison of four neural network architectures (MLP, ResNet, U-Net, FNO) formulated as autoregressive state-transition operators that predict full DFN internal states across a wide range of operating conditions. To ensure a controlled architectural comparison, all models are trained under a unified framework using multi-step unrolling and current-conditioning, isolating the impact of spatial inductive bias. Results demonstrate that the U-Net's multi-scale feature hierarchy achieves a mean final-step nRMSE of 3% averaged across all internal state variables after 300-step autoregressive rollouts, while providing a 5.38x speed-up over the numerical solver. These findings highlight spatial inductive bias as a critical determinant of surrogate performance, advancing the development of surrogates for internal state observability for next-generation battery management systems and digital twins.

Long-span roadway bridges exhibit highly localized structural responses under vehicular loading, making repeated FE analysis computationally expensive for applications such as influence surface generation and structural digital twins. Existing SciML approaches struggle to accurately capture these localized responses. To address this challenge, this study proposes an adaptive-trunk DeepONet for localized structural response prediction in large-scale bridge systems. The framework dynamically constructs a load-dependent learning domain using a KNN strategy, allowing the network to focus on structural influence zones. The trunk network is further enhanced using distance-aware features that encode the geometric relationship between the load and structural nodes. A physics-based full-field reconstruction is incorporated through a stiffness-informed Schur complement formulation, enabling predictions at adaptive nodes to be extended to the entire structural domain. To enable scalable training, response data are generated using a reduced-order equivalent shell model that preserves the dominant global behavior while significantly reducing computational cost. The proposed framework is validated on both a benchmark bridge model and the real-world Mussafah Bridge. Results show that the method achieves FEM-level accuracy with relative errors below 5%, while reducing the total response evaluation time (including full-field reconstruction) by approximately 60x; excluding the post-processing reconstruction step, the AD-DeepONet inference is up to four orders of magnitude faster than FEM. In addition, the framework enables rapid generation of full-field responses, influence lines, and influence surfaces under arbitrary vehicular loading configurations, demonstrating strong potential for large-scale bridge analysis and digital twin applications.

Artificial intelligence (AI) agents promise to accelerate drug discovery by compressing interpretation and decision-making loops, but practical deployment requires trusted evaluation on realistic program decisions. We introduce TherapeuticsBench Preclinical Pharmacology (TxBench-PP), a verifiable benchmark for small-molecule preclinical pharmacology and the first focused slice of a broader TherapeuticsBench effort across drug-discovery stages and therapeutic modalities. TxBench-PP tests whether agents can recover accurate conclusions from real-world assay data rather than memorized facts from literature. The benchmark contains 100 evaluations indexed by program stage, assay type, and task structure, spanning mechanism-of-action (MoA) and pharmacodynamic (PD) reasoning, compound-target engagement, causal target validation, developability and safety, and translational efficacy. Agents receive realistic workflow snapshots, inspect files in a coding environment, and return structured answers graded deterministically. Across 16 model-harness configurations, comprising 11 models and 4,800 trajectories, no system reliably recovered preclinical pharmacology decisions. The strongest configuration, Claude Opus 4.8 / Pi, passed 59.3\% of endpoint attempts (178/300; 95\% CI, 51.1-67.6), followed by GPT-5.5 / Pi at 55.3\% (166/300; 47.0-63.6).

Reliable benchmarking is critical for developing machine learning models for tandem mass spectrometry (MS/MS) based molecule discovery. Subtle issues in experimental design and model evaluation procedures can degrade the trustworthiness of such benchmarks and lead to erroneous conclusions. We conduct a thorough review of model evaluation issues in the recent MS/MS machine learning literature, using the standard MassSpecGym benchmark suite as a case study to illustrate the impact of these issues. We find evaluation issues in at least 17 of 26 papers reporting MassSpecGym benchmark results in the first year of its adoption. We isolate three classes of failures: (i) data leakage, (ii) shortcut learning, and (iii) implementation bugs and metric divergence. Through extensive experimentation and code replication, we quantify the impact of these issues and show how they corrupt the evaluation standards MassSpecGym was designed to enforce. We distill our findings into recommendations generalizable to MS/MS challenges, benchmarks, and custom evaluation setups. We also release MassSpecGym v1.5, an implementation of our recommendations in the MassSpecGym benchmarking suite which addresses the failure modes identified in this audit. MassSpecGym v1.5 is publicly available at https://github.com/pluskal-lab/MassSpecGym.

While geostrophy remains the simplest and most practical balance to extract velocity information from sea-surface height anomaly (SSHa), confusions remain within the oceanographic community to what extent this balance can be applied to altimetric observations with the launch of the Surface Water and Ocean Topography (SWOT) satellite. Given the limited temporal resolution of SWOT, many studies have resorted to claiming that the spatially filtered SSHa fields correspond to the geostrophic component. This introduces the ambiguity of which spatial scale to choose. Here, we build upon the recent developments in internal tide (IT) corrections (Yadidya et al., 2025) and apply a dynamic mode decomposition (DMD)-based method introduced by Lapo et al. (2025) to robustly extract the geostrophic component associated with sub-inertial frequencies from the SWOT one-day-repeat orbit; we distribute the global dataset as a public good. We provide the joint probability density function (PDF) of vorticity and strain, and spectra of SSHa at a few cross-over regions.

We introduce a novel active-learning framework for failure probability estimation in structural reliability analysis that integrates Active Kriging Monte Carlo simulation with conformal prediction. The proposed approach employs an adaptive cross-conformal strategy specifically designed for small-sample settings and kriging surrogate models using the J+GP conformal estimator. Unlike standard AK-MCS methods, the proposed framework provides distribution-free guarantees on prediction errors, leading to more reliable classification of samples near the limit-state surface. This improved uncertainty quantification enhances both the accuracy and robustness of failure probability estimates, especially for rare-event regimes where such efficiency is crucial. Reproducible numerical results illustrate the effectiveness of the method and also compare it to classical approaches on well-established benchmarks.

Factor models are popular approaches for analyzing high-dimensional data to extract low-rank signals and estimate covariances. They decompose the covariance matrix as the sum of low-rank and diagonal components. A key issue is how to choose the latent dimension $k$, which is particularly challenging when the factor model only holds approximately and in low signal-to-noise scenarios. Bayesian overfitted factor models specify an upper bound on $k$ and rely on structured shrinkage priors to effectively remove extra components. Such approaches are popular and effective, but computationally expensive. We propose a much faster \texttt{EigenBayes} approach that provides valid uncertainty quantification, based on spectral estimation of latent factors and adaptive empirical Bayes calibration of key hyperparameters. The resulting posterior distribution factorizes across outcomes and is analytically tractable, bypassing Markov chain Monte Carlo. We show that \texttt{EigenBayes} adapts to the signal-to-noise ratio of each outcome and latent dimension, while shrinking superfluous latent components to zero. We establish favorable asymptotic properties and demonstrate strong empirical performance in numerical experiments and a genomics application, where EigenBayes outperforms state-of-the-art alternatives.

Brain network is commonly divided into modules for analyzing their functionally segregated roles for group-level analysis in neuroimaging studies. Here, we introduce stochastic modules within brain networks for a robust probabilistic measurement of structural-functional module consistency (SFMC) in a group of subjects. Specifically, a stochastic module can be regarded as the chance of a brain region across subjects potentially being assigned to a group-level sub-network, characterized as an assignment probability for this brain region. This novel method has two advantages for evaluating inhomogeneous modules in brain networks. The first is that it can robustly evaluate the consistency between brain structural and functional modules whose population sizes are not necessary the same, and the second is that it is able to take into account the inter-individual variability of the modules for the groups. Moreover, compared with the conventional structural-functional coupling approach, our stochastic module-based method reveals a more pronounced decline in the coupling between structure and function, indicating stronger developmental reorganization. Our results using the dataset from Baby Connectome Project (BCP) show that the SFMC decreases from 0 to 5 years old, and is greater in primary brain regions, such as visual areas, while lower in more advanced cognitive regions, including those related to attention, control, and default mode network.